Discussion Speculation: Zen 4 (EPYC 4 "Genoa", Ryzen 7000, etc.)

[Vattila]

Except for the details about the improvements in the microarchitecture, we now know pretty well what to expect with Zen 3.

The leaked presentation by AMD Senior Manager Martin Hilgeman shows that EPYC 3 "Milan" will, as promised and expected, reuse the current platform (SP3), and the system architecture and packaging looks to be the same, with the same 9-die chiplet design and the same maximum core and thread-count (no SMT-4, contrary to rumour). The biggest change revealed so far is the enlargement of the compute complex from 4 cores to 8 cores, all sharing a larger L3 cache ("32+ MB", likely to double to 64 MB, I think).

Hilgeman's slides did also show that EPYC 4 "Genoa" is in the definition phase (or was at the time of the presentation in September, at least), and will come with a new platform (SP5), with new memory support (likely DDR5).

What else do you think we will see with Zen 4? PCI-Express 5 support? Increased core-count? 4-way SMT? New packaging (interposer, 2.5D, 3D)? Integrated memory on package (HBM)?

Vote in the poll and share your thoughts!

 

[Racan]

5ghz boost for 7800X3D while disappointing is still 500mhz over the 5800X3D and according to TPW the 13900K is only 6.2% faster than the 5800X3D at 1080p with an RTX 4090 in their 53 game benchmark. https://www.techpowerup.com/review/rtx-4090-53-games-core-i9-13900k-vs-ryzen-7-5800x3d/2.html

 

[BorisTheBlade82]

Part 3 of the CnC Zen4 series:

AMD’s Zen 4, Part 3: System Level Stuff, and iGPU

We covered Zen 4’s core architecture in depth in two articles. This one will focus on anything we didn’t manage to get to. Some of these details may be specific to the particular CPU sa…

chipsandcheese.com

This again confirms the IFoP bottlenecking RAM bandwidth - it was a questionable decision by AMD not to use wide-mode for single CCD SKUs.

 

[biostud]

5ghz boost for 7800X3D while disappointing is still 500mhz over the 5800X3D and according to TPW the 13900K is only 6.2% faster than the 5800X3D at 1080p with an RTX 4090 in their 53 game benchmark. https://www.techpowerup.com/review/rtx-4090-53-games-core-i9-13900k-vs-ryzen-7-5800x3d/2.html

Will definitely wait for review now to decide which way to go. I did not expect such a huge reduction in frequency. Now it comes down to price and minimum fps. I will be GPU limited in anything I game, so I don't expect any difference on my system when it comes to max fps.

 

[Harry_Wild]

AMD Expands Desktop Ryzen 7000 "Zen 4" Processor Family with 65W Models

Besides the sensational Ryzen 7000X3D processors taking the fight to Intel's "Raptor Lake," AMD expanded the desktop Ryzen 7000 even downwards, with the introduction of three new 65 W processor SKUs that include boxed stock coolers. These include the Ryzen 5 7600 6-core/12-thread, the Ryzen 7...

www.techpowerup.com

AMD Expands Desktop Ryzen 7000 "Zen 4" Processor Family with 65W Models
by btarunr Yesterday, 20:31 Discuss (7 Comments)
Besides the sensational Ryzen 7000X3D processors taking the fight to Intel's "Raptor Lake," AMD expanded the desktop Ryzen 7000 even downwards, with the introduction of three new 65 W processor SKUs that include boxed stock coolers. These include the Ryzen 5 7600 6-core/12-thread, the Ryzen 7 7700 8-core/16-thread, and the Ryzen 9 7900 12-core/24-thread. There's no 16-core part in this segment. These processors come with TDP values set at just 65 W, and PPT values in the range of 90 W to 120 W, and so their clock speeds and maximum boost speeds are lower compared to the 7000X series, with more aggressive power-management.

The 7600 boosts up to 5.10 GHz, and packs a 65 W-capable Wraith Stealth boxed cooling solution. The 7700 boosts up to 5.30 GHz, and the 7900 up to 5.40 GHz. Both the 7700 and 7900 include a Wraith Prism RGB cooler that can handle thermal loads of up to 140 W. The three chips are priced lower than their 7000X series cousins, with the 7600 going for USD $229, the 7700 at $329, and the 7900 at $429. The three chips are drop-in compatible with existing Socket AM5 motherboards without needing any BIOS update.

 

[eek2121]

Anyone who buys old tech thinking it's new. I.e. anyone who doesn't follow this stuff as closely as we do.

The average consumer would definitely notice the difference in performance, features, battery life, etc.

As for Raptor Lake, it's looking like only the desktop is true rebrands, but either way, at least higher number == better, and there're no significant feature differences. Of course, still not ideal, but not nearly this bad.

2023 Zen 2 is not OG Zen 2, so technically it IS a new product. The newer Zen 2 part will likely be faster thanks to LPDDR5 and possibly better clocks.

EDIT: looking back at the slides, it appears AMD is gimping the Zen 2 parts by giving them very little cache. Time will tell how that plays out.

Suffice to say, as long as a bigger model number is always faster, the naming scheme should not be confusing.

 

[Timmah!]

ofc it does not make sense, its almost 100 percent fake.

I stand corrected, it was actually genuine. Surprised pikachu face.

That said, personally i am rather disappointed. It clearly comes with clock penalty once again, as seen on lower base clock and 5GHz top boost clock of 7800X3D. There will be no doubt lower all-core performance in the likes of Cinebench, Blender, etc... because of this.
Since the cache is only on the primary die, they can still claim that 5,7/5,6GHz boost on dual die models, but i have to wonder, when exactly will those cores boost to that frequency, since i presume in lower core count loads, when those frequencies are attainable, usually the cores on the primary chip are used. What if you run some app, that does not benefit from bigger cache, but would from higher clocks - how will the CPU know to run it on the non-vcache die to get that higher frequency? Unless you force it manually via Process Lasso or something. This looks pretty similar to Intel big.little situation, even if the cores are the same.

Ultimately, i need to see reviews, because the info we were given is wildly insufficient, too many questions. I still could return my 7950x till the end of January and get 7950x3D in February instead (would mean keeping everything in boxes for another month or 2), but i honestly dont know at this point, whether that would good idea. And the reviews probably wont come before February to help me decide.

 

[nicalandia]

Some info on Bergamo...

 

[Kocicak]

... the info we were given is wildly insufficient, too many questions ...

We know clocks of the one chiplet CPU, we know, that they could not put cache chips on both chiplets to maintain lower thread count performance. We also know that 7800X3D is just 15% better than 5800X3D and most likely worse than 13900K in gaming. We also know that 7950X3D is probably somewhat better than 13900K in gaming.

I think we already know almost everything we need to know.

I do not see any crushing victory of the new AMD 3D chips, 7800X3D will be competitive with Intel in gaming only when you factor in the price of the CPUs, 7950X3D will probably trade blows with 13900KS, and if it still wins, only by a few percent.

 

[inf64]

We know clocks of the one chiplet CPU, we know, that they could not put cache chips on both chiplets to maintain lower thread count performance. We also know that 7800X3D is just 15% better than 5800X3D and most likely worse than 13900K in gaming. We also know that 7950X3D is probably somewhat better than 13900K in gaming.

I think we already know almost everything we need to know.

I do not see any crushing victory of the new AMD 3D chips, 7800X3D will be competitive with Intel in gaming only when you factor in the price of the CPUs, 7950X3D will probably trade blows with 13900KS, and if it still wins, only by a few percent.

13900K is just 6.1% faster than 5800X3D @ 1080p using the 4090 : https://www.techpowerup.com/review/rtx-4090-53-games-core-i9-13900k-vs-ryzen-7-5800x3d/2.html

7800X3D being 15% faster than 5800X3D should put it clearly ahead of 13900K and most likely the new 13900KS. Then, we have 7950X3D that will likely have a better boost on the 3D cache chiplet vs 7800X3D, so it should perform even better. So all in all, I'm not sharing the same opinion that 7950X3D " will probably trade blows with 13900KS, and if it still wins, only by a few percent.". It should be clearly a better gaming chip with decent performance advantage.

 

[nicalandia]

Okay, so The EPYC 9754 Bergamo is a 128 Core and 256 Thread CPU, some people thought that SMT was going to be disabled.

 

[BorisTheBlade82]

It is said to be having 10 dies - maybe 2 sIOD.

https://twitter.com/x/status/1610923017869340672

 

[Kocicak]

7800X3D being 15% faster than 5800X3D should put it clearly ahead of 13900K and most likely the new 13900KS.

Why AMD did not compare 7800X3D to 13900K then? I mean what better message could they send than that?

 

[Timmah!]

13900K is just 6.1% faster than 5800X3D @ 1080p using the 4090 : https://www.techpowerup.com/review/rtx-4090-53-games-core-i9-13900k-vs-ryzen-7-5800x3d/2.html

7800X3D being 15% faster than 5800X3D should put it clearly ahead of 13900K and most likely the new 13900KS. Then, we have 7950X3D that will likely have a better boost on the 3D cache chiplet vs 7800X3D, so it should perform even better. So all in all, I'm not sharing the same opinion that 7950X3D " will probably trade blows with 13900KS, and if it still wins, only by a few percent.". It should be clearly a better gaming chip with decent performance advantage.

I dont think this is neccesarily the case. Actually, its IMO more likely that the 3d cache chiplet of 7950x3d will have the same boost as 7800x3d.

 

[Abwx]

Why AMD did not compare 7800X3D to 13900K then? I mean what better message could they send than that?

Because they did put the lights preferably on the 7950X3D, otherwise there s slides that allow to have an idea of the 7800X3D perfs.

AMD Ryzen 7000X3D: 7950X3D, 7900X3D & 7800X3D mit hohem Takt im Februar

Es ist offiziell: AMD Ryzen 7000X3D erscheint im Februar. 7950X3D, 7900X3D & 7800X3D bieten 3D V-Cache und einen hohen Takt bei 120 W TDP.

www.computerbase.de

 

[poke01]

We know clocks of the one chiplet CPU, we know, that they could not put cache chips on both chiplets to maintain lower thread count performance. We also know that 7800X3D is just 15% better than 5800X3D and most likely worse than 13900K in gaming. We also know that 7950X3D is probably somewhat better than 13900K in gaming.

I think we already know almost everything we need to know.

I do not see any crushing victory of the new AMD 3D chips, 7800X3D will be competitive with Intel in gaming only when you factor in the price of the CPUs, 7950X3D will probably trade blows with 13900KS, and if it still wins, only by a few percent.

Are people forgetting efficiency?

If Intel needs 350watts to achieve that then that's not a win!

 

[Timmah!]

We know clocks of the one chiplet CPU, we know, that they could not put cache chips on both chiplets to maintain lower thread count performance. We also know that 7800X3D is just 15% better than 5800X3D and most likely worse than 13900K in gaming. We also know that 7950X3D is probably somewhat better than 13900K in gaming.

I think we already know almost everything we need to know.

I do not see any crushing victory of the new AMD 3D chips, 7800X3D will be competitive with Intel in gaming only when you factor in the price of the CPUs, 7950X3D will probably trade blows with 13900KS, and if it still wins, only by a few percent.

As i said, we dont know what will happen when you will want to run app that is more frequency than cache-sensitive. Will it run on the second non-cache die? Cause if not, that 5,7GHz is boost is there just a hypothetical number, in reality you wont never see it - cause low-threaded stuff will run on primary v-cache die, all core loads you wont see 5,7GHz (i presume).
Additionally, if the second non-vcache die can boost to 5,7, does it mean its of higher quality than the second die in vanilla 7950x? Can all the cores in 7950x boost to that speed, or just bunch of the best ones (which are then by default chosen to be loaded in lower-threaded load scenarios.)

Yes, we can infer that all the CPUs will be probably within 10~15 percent on average of each other, depending on the usage, so in a way, whichever you get, you will be more than OK, its not like there will be 50 percent differences. Still, if i am interested in details, or if i really want to choose the perfect CPU for me, there are indeed questions unanswered.

 

[KompuKare]

I do not see any crushing victory of the new AMD 3D chips, 7800X3D will be competitive with Intel in gaming only when you factor in the price of the CPUs, 7950X3D will probably trade blows with 13900KS, and if it still wins, only by a few percent.

Another question is, will 7800X3D perform well with cheaper memory like 5800X3D does?

Not everyone wants to buy the really expensive top DDR5 nor tune it. That was a big advantage of the current 5800X3D (that and AM4 compatibility).

For the dual CCD 3D chips; well they probably do require fast memory or else what happens to threads on the 3D cacheless CCD when they have to go to main memory.

Both Intel and AMD really need some easy to use per-application config file with presets where they provide the best defaults for their CPUs in terms of which type of core (in Intel's case) or which CCD (in AMD's case) a program and it's threads should run on.

 

[Kocicak]

Because they did put the lights preferably on the 7950X3D, otherwise there s slides that allow to have an idea of the 7800X3D perfs.

This slide is different than what was shown in presentation.

 

[inf64]

Why AMD did not compare 7800X3D to 13900K then? I mean what better message could they send than that?

Because they compared 8C previous gen vs 8C next gen to show the gains? I have no doubt that 7800X3D will be faster gaming chip than 13900K(S).

This slide is different than what was shown in presentation.

Different set of games in the two slides.

I dont think this is neccesarily the case. Actually, its IMO more likely that the 3d cache chiplet of 7950x3d will have the same boost as 7800x3d.

I doubt that, but let's see. 5800X3d had a max boost clock penalty of ~5% vs 5800X. For 7800X3D it's 7.4% vs 7700X. If similar is true for 7950X3D, 7950X3D could have ~5.2Ghz boost on the Vcache chiplet.

 

[Kocicak]

As i said, we dont know what will happen when you will want to run app that is more frequency than cache-sensitive. Will it run on the second non-cache die? Cause if not, that 5,7GHz is boost is there just a hypothetical number, in reality you wont never see it - cause low-threaded stuff will run on primary v-cache die, all core loads you wont see 5,7GHz (i presume).
Additionally, if the second non-vcache die can boost to 5,7, does it mean its of higher quality than the second die in vanilla 7950x? Can all the cores in 7950x boost to that speed, or just bunch of the best ones (which are then by default chosen to be loaded in lower-threaded load scenarios.)

I believe that the threads must be primarily assigned to die with cache for game threads and to bare die with other application, this does not seem difficult to do for me, but I know nothing about inner workings of an operating system, I wonder if AMD is working on some Windows update with Microsoft.

 

[inf64]

I believe that the threads must be primarily assigned to die with cache for game threads and to bare die with other application, this does not seem difficult to do for me, but I know nothing about inner workings of an operating system, I wonder if AMD is working on some Windows update with Microsoft.

They were working with various software vendors, Microsoft included. Here is the PCworld video with more details:

 

[Timmah!]

I doubt that, but let's see. 5800X3d had a max boost clock penalty of ~5% vs 5800X. For 7800X3D it's 7.4% vs 7700X. If similar is true for 7950X3D, 7950X3D could have ~5.2Ghz boost on the Vcache chiplet.

We dont know if the 5GHz boost on 7800x3d is cause of product segmentation or because v-cache related limitation. If its cause of v-cache, i think its pretty possible same limitation will concern 7950x as well.
But as you say, we shall see. I am curious, how it all plays out, no doubt.

 

[Abwx]

This slide is different than what was shown in presentation.

Whatever, according to Computerbase games tests, wich are somewhat favourable to Intel, the 13900K has 15% better perf than the 5800X3D.

So with 20% improvement the 7800X3D should be faster than the former, and that s with a 5GHz limited chip, that s quite low given that at 5.4GHz a single 7700X core use only 28W in Cinebench wich is a more demanding load than games, possibly that they want the 7900X3D/7950X3D to look better than the lower priced part.

 

[Timmah!]

I believe that the threads must be primarily assigned to die with cache for game threads and to bare die with other application, this does not seem difficult to do for me, but I know nothing about inner workings of an operating system, I wonder if AMD is working on some Windows update with Microsoft.

But does not operating system assign the threads? And Windows is historically pretty lousy at that. It may sound pretty straightforward in theory, but it can be completely different thing in practice. I would be bit cautious here.

 

[In2Photos]

They were working with various software vendors, Microsoft included. Here is the PCworld video with more details:

Thanks for the video. Answered some of the questions we have been asking. The fact that they didn't give a price is odd. I hope they are not waiting to see the public's response to determine where to price it.